WO2017088470A1 - 图像分类方法及装置 - Google Patents
图像分类方法及装置 Download PDFInfo
- Publication number
- WO2017088470A1 WO2017088470A1 PCT/CN2016/087562 CN2016087562W WO2017088470A1 WO 2017088470 A1 WO2017088470 A1 WO 2017088470A1 CN 2016087562 W CN2016087562 W CN 2016087562W WO 2017088470 A1 WO2017088470 A1 WO 2017088470A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- images
- classified
- threshold
- merge threshold
- information
- Prior art date
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/285—Selection of pattern recognition techniques, e.g. of classifiers in a multi-classifier system
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/161—Detection; Localisation; Normalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/211—Selection of the most significant subset of features
- G06F18/2115—Selection of the most significant subset of features by evaluating different subsets according to an optimisation criterion, e.g. class separability, forward selection or backward elimination
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/762—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using clustering, e.g. of similar faces in social networks
- G06V10/763—Non-hierarchical techniques, e.g. based on statistics of modelling distributions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/172—Classification, e.g. identification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23211—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with adaptive number of clusters
Definitions
- the present disclosure relates to the field of image recognition technologies, and in particular, to an image classification method and apparatus.
- smart terminals represented by smart phones can realize more and more functions.
- a user can take a self-portrait or take a photo of another person as needed, or download a photo from himself or another person's space to the local storage.
- the present disclosure provides an image classification method and apparatus.
- an image classification method including:
- the face pose information in the two images to be classified are respectively determined according to the pre-established face pose classifier model
- the face pose information in the two images to be classified are respectively determined, thereby determining the between the two images to be classified.
- the target cluster merges the thresholds, and then the two images to be classified are classified according to the target clustering merge threshold.
- it also includes:
- the accuracy of the two images to be classified can be further improved by the embodiment provided by the present disclosure. If the two images to be classified are smaller than a certain degree of similarity, that is, when it is determined from the two images to be classified that they do not obviously belong to the same type of image, the method provided by the present disclosure is not required. Therefore, at least a certain degree of similarity between the two images to be classified can be satisfied.
- the determining a target clustering merge threshold between the two to-categorized images includes:
- the preset cluster merge threshold is increased in a preset manner, and the enlarged preset cluster is The merge threshold is used as the target cluster merge threshold.
- the face pose information in the two images to be classified are all large-angle pose information, it indicates that the face features in the two images to be classified are rich, and the clustering threshold between the two needs to be increased to improve the accuracy of the classification. degree.
- the determining a target clustering merge threshold between the two to-categorized images includes:
- the preset cluster merge threshold is decreased according to a preset manner, and the reduced preset cluster merge threshold is used as the threshold Target cluster merge threshold.
- the face pose information in the two images to be classified are different angle pose information, it indicates that the content of the face features in the two images to be classified is not rich, and the cluster merge threshold between the two needs to be lowered to improve The accuracy of the classification.
- the determining a target clustering merge threshold between the two to-categorized images includes:
- the preset cluster merge threshold is used as the target cluster merge threshold.
- the preset cluster merge threshold may be directly used as the target cluster merge threshold.
- the classifying the two to-be-categorized images by using the determined target clustering merge threshold including:
- the two to-categorized images are taken as the same type of image
- the two to-categorized images are treated as different types of images.
- an image classification apparatus comprising:
- a face pose information determining module configured to determine a face of the two to-be-categorized images according to a pre-established face pose classifier model when acquiring two to-be-categorized images each containing face feature information Gesture information
- a target clustering merge threshold determining module configured to determine a target clustering merge threshold between the two to-be-categorized images according to the face pose information in the two to-be-classified images
- an image classification module configured to classify the two images to be classified by using the determined target clustering merge threshold.
- it also includes:
- a similarity calculation module configured to calculate a similarity between the two images to be classified
- the threshold determining module is configured to determine whether the similarity is greater than a preset threshold.
- the target clustering merge threshold determining module includes:
- a first clustering merge threshold obtaining sub-module configured to acquire a preset clustering merge threshold
- a large angle attitude information judging sub-module configured to determine whether the facial gesture information in the two to-be-classified images are all large-angle attitude information
- a first target clustering merge threshold sub-module configured to increase the preset clustering merge threshold according to a preset manner when the face pose information in the two to-be-classified images are all large-angle pose information
- the increased preset clustering merge threshold is used as a target clustering merge threshold.
- the target clustering merge threshold determining module includes:
- a second clustering merge threshold acquisition sub-module configured to acquire a preset cluster merge threshold
- a different angle attitude information determining sub-module configured to determine whether the face pose information in the two to-be-classified images are respectively different angle posture information
- a second target clustering and merging threshold sub-module wherein the facial gesture information in the two to-be-classified images is different angle posture information, and the preset cluster merging threshold is reduced according to a preset manner, and the threshold is reduced.
- the preset clustering merge threshold is used as the target clustering merge threshold.
- the target clustering merge threshold determining module includes:
- a third cluster merge threshold acquisition sub-module configured to acquire a preset cluster merge threshold
- a front posture information determining sub-module configured to determine whether the face posture information in the two to-be-classified images are all front posture information
- the third target clustering and merging threshold sub-module is configured to use the preset clustering merging threshold as the target clustering merging threshold when the facial gesture information in the two to-be-classified images is the frontal posture information.
- the image classification module includes:
- a clustering combined value calculation sub-module configured to calculate a clustered combined value of the two images to be classified
- a clustering value judgment sub-module configured to determine whether the cluster merge value is greater than the target cluster merge threshold
- the same type of image determining sub-module configured to use the two images to be classified as the same type of image when the clustering combined value is greater than the target clustering combining threshold;
- the different types of image determining sub-modules are configured to use the two images to be classified as different types of images when the cluster combining value is not greater than the target clustering combining threshold.
- a terminal including:
- a memory for storing processor executable instructions
- processor is configured to:
- the face pose information in the two images to be classified are respectively determined according to the pre-established face pose classifier model
- the image classification method and apparatus determine the face pose information in the two images to be classified according to the pre-established arbitrary light pose classifier model when classifying the two images to be classified. And determining a target clustering merge threshold between the two images to be classified, and then classifying the two images to be classified according to the target clustering merge threshold.
- FIG. 1 is a flowchart of an image classification method according to an exemplary embodiment
- FIG. 2 is a flowchart of an image classification method according to another exemplary embodiment
- FIG. 3 is a flow chart of step S120 of Figure 1;
- FIG 4 is another flow chart of step S120 of Figure 1;
- FIG. 5 is still another flow chart of step S120 of Figure 1;
- Figure 6 is a flow chart of step S130 of Figure 1;
- FIG. 7 is a schematic diagram of an image classification apparatus according to an exemplary embodiment
- FIG. 8 is a schematic diagram of an image classification apparatus according to still another exemplary embodiment.
- FIG. 9 is a schematic diagram of a target clustering merge threshold determination module of FIG. 7;
- FIG. 10 is another schematic diagram of the target clustering merge threshold determining module of FIG. 7; FIG.
- FIG. 11 is another schematic diagram of the target clustering merge threshold determination module of FIG. 7;
- Figure 12 is a schematic diagram of the image classification module of Figure 7;
- FIG. 13 is a block diagram of a terminal, according to an exemplary embodiment.
- the embodiment of the present disclosure first provides an image classification method, as shown in FIG. 1 , which may include the following steps:
- step S110 when the two images to be classified that all include the face feature information are acquired, the face pose information in the two images to be classified is respectively determined according to the pre-established face pose classifier model.
- the face pose information in the two images to be classified may be determined by a pre-established face pose classifier model.
- a pre-established face pose classifier model Such as: face, side face, left face, face, or two positive face images.
- step S120 a target clustering merge threshold between the two images to be classified is determined according to the face pose information in the two images to be classified.
- the face pose classifier model established in advance, it is also necessary to acquire the corresponding cluster according to the face pose information. Merge thresholds.
- An image photo may be a front face gesture information or a side face gesture information of other angles.
- two image photos may be two positive face image photos, one front face and one side face, two side faces, etc., each case corresponding to a cluster merge threshold, wherein the side
- the face pose information may include side face pose information at different angles.
- the cluster merge threshold between the two images to be classified may be determined as a target cluster merge threshold by means of table lookup or automatic generation.
- step S130 the two images to be classified are classified by using the determined target clustering merge threshold.
- the embodiment of the present disclosure further discriminates two images to be classified whose similarity is greater than a set threshold, the accuracy of the classification is further improved, so when the clustering value of the two images to be classified is greater than the target clustering threshold , indicating that the two images to be classified belong to one category, otherwise they do not belong to one category.
- the image classification method provided in the embodiment of the present disclosure determines the face pose information in the two images to be classified according to the pre-established arbitrary light pose classifier model when classifying the two images to be classified. Determining a target clustering merge threshold between the two images to be classified, and then classifying the two images to be classified according to the target clustering merge threshold. By determining the face pose information, it can be more accurately determined whether the two images to be classified belong to the same type of image, thereby improving the image classification efficiency.
- the method may further include the following steps:
- step S140 the similarity between the two images to be classified is calculated.
- step S150 it is determined whether the similarity is greater than a preset threshold.
- step S110 is performed.
- the embodiment of the present disclosure mainly classifies two images with certain similarities, that is, when it is not possible to distinguish whether the two are the same type of images by similarity, the two images to be classified can be further improved by using the embodiment provided by the present disclosure. Accuracy. If the two images to be classified are smaller than a certain degree of similarity, that is, when it is determined from the two images to be classified that they do not obviously belong to the same type of image, the method provided by the present disclosure is not required. Therefore, at least a certain degree of similarity between the two images to be classified can be satisfied.
- step S120 may further include the following steps:
- step S121 a preset cluster merge threshold is acquired.
- step S122 it is determined whether the face pose information in the two images to be classified are all large angle pose information.
- step S123 the preset cluster merge threshold is increased according to a preset manner, and the increased preset cluster merge threshold is used as the threshold Target cluster merge threshold.
- the face pose information in the two images to be classified are all large-angle pose information, it indicates that the face features in the two images to be classified are rich, and the clustering threshold between the two needs to be increased to improve the accuracy of the classification. degree.
- step S120 may further include the following steps:
- step S124 a preset cluster merge threshold is acquired.
- step S125 it is determined whether the face pose information in the two images to be classified are respectively different angle pose information.
- step S126 the preset cluster merge threshold is decreased according to the preset manner, and the reduced preset cluster merge threshold is used as the target. Cluster merge thresholds.
- the face pose information in the two images to be classified are different angle pose information, it indicates that the content of the face features in the two images to be classified is not rich, and the cluster merge threshold between the two needs to be lowered to improve The accuracy of the classification.
- step S120 may further include the following steps:
- step S127 a preset cluster merge threshold is acquired.
- step S1208 it is determined whether the face pose information in the two images to be classified are all front posture information.
- step S129 the preset cluster merge threshold is used as the target cluster merge threshold.
- the preset cluster merge threshold may be directly used as the target cluster merge threshold.
- facial gesture information is described in detail, that is, both the large angle posture information, the different angle posture information, and the front posture information.
- the face feature information in the image to be classified needs to be extracted, through the person
- the face feature information is used to determine face pose information in the image to be classified.
- a three-dimensional coordinate system in a horizontal, vertical, and vertical direction may be established on a frontal face, and feature information such as a human eye, a nose, an ear, and the like in the image may be extracted to determine a face pose of the image to be classified.
- the amount, such as the face pose information in the image to be classified, is 20 degrees to the left.
- a face pose sample library can be established by collecting a plurality of face pose sample image photos.
- the face pose sample library may include a plurality of photo images of a plurality of face poses.
- the face pose sample library includes: 1000 face images of frontal face gestures, 1000 photos of 10 degree face gestures, 1000 images of 20 degree face gestures, and the like.
- the 10 degree face gesture photo may be an angle between the front of the face and the lens when shooting. Among them, the different angles and the number of photos can be set according to actual needs.
- the face photos contained in the face pose sample library should include men and women. Photo images, such as photo images that can include old people and children, and so on.
- the existing facet can be used to train the photo image in the established face pose sample library to obtain a face pose classifier model.
- a non-linear classifier can be used to train a face pose classifier model for image photographs in a face pose sample library.
- a SVM Small Vector Machine
- CNN Convolutional Neural Networks
- step S130 may further include the following steps:
- step S131 cluster combination values of two images to be classified are calculated.
- step S132 it is determined whether the cluster merge value is greater than the target cluster merge threshold.
- step S133 the two images to be classified are taken as the same type of image.
- step S134 the two images to be classified are taken as different types of images.
- the image classification method provided in the embodiment of the present disclosure determines the face pose information in the two images to be classified according to the pre-established arbitrary light pose classifier model when classifying the two images to be classified. Determining a target clustering merge threshold between the two images to be classified, and then classifying the two images to be classified according to the target clustering merge threshold. By determining the face pose information, it can be more accurately determined whether the two images to be classified belong to the same type of image, thereby improving the image classification efficiency.
- portions of the technical solution of the present disclosure that contribute substantially or to the prior art may be embodied in the form of a software product stored in a storage medium, including a plurality of instructions for causing a A computer device (which may be a personal computer, server, or network device, etc.) performs all or part of the steps of the methods described in various embodiments of the present disclosure.
- the foregoing storage medium includes various types of media that can store program codes, such as a read only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.
- an embodiment of the present disclosure further provides an image classification device, which is located in a terminal. As shown in FIG. 7, the device includes: a face pose information determining module 10, and a target group. Class merge threshold determination module 20 and image classification module 30, wherein
- the face pose information determining module 10 is configured to determine the people in the two images to be classified according to the pre-established face pose classifier model when acquiring the two images to be classified that all include the face feature information Face gesture information;
- the face pose information in the two images to be classified may be determined by a pre-established face pose classifier model.
- a pre-established face pose classifier model Such as: face, side face, left face, face, or two positive face images.
- the target clustering merge threshold determining module 20 is configured to determine a target clustering merge threshold between the two to-categorized images according to the face pose information in the two to-be-classified images;
- the face pose classifier model established in advance, it is also necessary to acquire the corresponding cluster according to the face pose information. Merge thresholds.
- An image photo may be a front face gesture information or a side face gesture information of other angles.
- two image photos may be two positive face image photos, one front face and one side face, two side faces, etc., each case corresponding to a cluster merge threshold, wherein the side
- the face pose information may include side face pose information at different angles.
- the cluster merge threshold between the two images to be classified may be determined as a target cluster merge threshold by means of table lookup or automatic generation.
- the image classification module 30 is configured to use the determined target clustering merge threshold to the two to be classified The images are sorted.
- the embodiment of the present disclosure further discriminates two images to be classified whose similarity is greater than a set threshold, the accuracy of the classification is further improved, so when the clustering value of the two images to be classified is greater than the target clustering threshold , indicating that the two images to be classified belong to one category, otherwise they do not belong to one category.
- the image classification device determines the face pose information in the two images to be classified according to the pre-established arbitrary light gesture classifier model when classifying the two images to be classified. Determining a target clustering merge threshold between the two images to be classified, and then classifying the two images to be classified according to the target clustering merge threshold. By determining the face pose information, it can be more accurately determined whether the two images to be classified belong to the same type of image, thereby improving the image classification efficiency.
- the apparatus further includes: a similarity calculation module 40 and a threshold determination module 50, wherein
- the similarity calculation module 40 is configured to calculate a similarity between the two images to be classified
- the threshold determination module 50 is configured to determine whether the similarity is greater than a preset threshold.
- the embodiment of the present disclosure mainly classifies two images with certain similarities, that is, when it is not possible to distinguish whether the two are the same type of images by similarity, the two images to be classified can be further improved by using the embodiment provided by the present disclosure. Accuracy. If the two images to be classified are smaller than a certain degree of similarity, that is, when it is determined from the two images to be classified that they do not obviously belong to the same type of image, the method provided by the present disclosure is not required. Therefore, at least a certain degree of similarity between the two images to be classified can be satisfied.
- the target clustering merge threshold determining module 20 includes: a first clustering merge threshold acquisition sub-module 21, and a large-angle attitude information determining sub-module. 22 and a first target clustering merge threshold sub-module 23, wherein
- the first clustering merge threshold acquisition sub-module 21 is configured to acquire a preset cluster merge threshold
- the large-angle attitude information judging sub-module 22 is configured to determine whether the face pose information in the two to-be-classified images are all large-angle pose information
- the first target clustering and merging threshold sub-module 23 is configured to increase the preset clustering merging threshold in a preset manner when the facial gesture information in the two to-be-classified images is the large-angle attitude information.
- the increased preset clustering merge threshold is used as a target clustering merge threshold.
- the face pose information in the two images to be classified are all large-angle pose information, it indicates that the face features in the two images to be classified are rich, and the clustering threshold between the two needs to be increased to improve the accuracy of the classification. degree.
- the target clustering merge threshold determining module 20 includes: a second clustering merge threshold acquiring sub-module 24, and determining different angle posture information. Sub-module 25 and second target cluster merge threshold sub-module 26, wherein
- the second clustering merge threshold acquisition sub-module 24 is configured to acquire a preset cluster merge threshold
- the different angle posture information judging sub-module 25 is configured to determine whether the face pose information in the two to-be-classified images are respectively different angle pose information
- the second target clustering and combining threshold sub-module 26 is configured to: the face pose information in the two to-be-classified images are respectively different angle pose information, and the preset cluster merge threshold is reduced according to a preset manner, The reduced preset cluster merge threshold is used as a target cluster merge threshold.
- the face pose information in the two images to be classified are different angle pose information, it indicates that the content of the face features in the two images to be classified is not rich, and the cluster merge threshold between the two needs to be lowered to improve The accuracy of the classification.
- the target clustering merge threshold determining module 20 includes: a third clustering merge threshold acquisition sub-module 27, a front posture information judging Module 28 and third target clustering merge threshold sub-module 29, wherein
- the third clustering merge threshold acquisition sub-module 27 is configured to acquire a preset cluster merge threshold
- the front posture information judging sub-module 28 is configured to determine whether the face pose information in the two to-be-classified images are all front posture information
- the third target clustering merge threshold sub-module 29 is configured to use the preset clustering merge threshold as the target clustering merge threshold when the face pose information in the two to-be-classified images are both front pose information.
- the preset cluster merge threshold may be directly used as the target cluster merge threshold.
- the image classification module 30 includes: a cluster combination value calculation sub-module 31, a cluster combination value judgment sub-module 32, and the same class.
- the cluster merge value calculation sub-module 31 is configured to calculate a cluster merge value of the two images to be classified;
- the cluster merge value judgment sub-module 32 is configured to determine whether the cluster merge value is greater than the target cluster merge threshold
- the same type of image determining sub-module 33 is configured to use the two images to be classified as the same type of image when the clustering combined value is greater than the target clustering combining threshold;
- the different class image determining sub-module 34 is configured to treat the two images to be classified as different types of images when the cluster combining value is not greater than the target clustering merge threshold.
- the image classification device obtaineds the posture information of the two faces to be classified when the two similarly classified images to be classified are classified, and the two similarities are determined by judging The pose information of the image to be classified determines a cluster merge threshold corresponding to the pose information of the two images to be classified.
- a cluster merge threshold corresponding to the pose information of the two images to be classified.
- the cluster merge threshold corresponding to the pose information of the two images to be classified is variable. If the attitude information of the two images to be classified are relatively close, for example, both When the angle is a large angle, the clustering threshold may be adjusted to a higher degree; if the posture information of the two images to be classified is different, such as one is the left face posture information, and the other is the right face posture information. At this time, the clustering between the two may be adjusted to a lower threshold; if the posture information of the two images to be classified belong to the frontal face posture information, the clustering threshold adjustment between the two may not be performed. Adjustment. That is, the clustering merge threshold corresponding to the posture information of the two images to be classified can be adaptive, and the two images to be classified can be accurately determined whether it is a photo image of the same person.
- FIG. 13 is a schematic structural diagram of an apparatus 1300 for image classification, according to an exemplary embodiment.
- device 1300 can be a mobile phone, a computer, a digital broadcast terminal, a messaging device, a gaming console, a tablet device, a medical device, a fitness device, a personal digital assistant, and the like.
- apparatus 1300 can include one or more of the following components: processing component 1302, memory 1304, power component 1306, multimedia component 1313, audio component 1310, input/output (I/O) interface 1312, sensor component 1314, And a communication component 1316.
- Processing component 1302 typically controls the overall operation of device 1300, such as operations associated with display, telephone calls, data communications, camera operations, and recording operations.
- Processing component 1302 can include one or more processors 1320 to execute instructions to perform all or part of the steps described above.
- processing component 1302 can include one or more modules to facilitate interaction between component 1302 and other components.
- processing component 1302 can include a multimedia module to facilitate interaction between multimedia component 1313 and processing component 1302.
- Memory 1304 is configured to store various types of data to support operation at device 1300. Examples of such data include instructions for any application or method operating on device 1300, contact data, phone book data, messages, pictures, videos, and the like.
- Memory 1304 can be implemented by any type of volatile or non-volatile storage device, or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read only memory (EEPROM), erasable Programmable Read Only Memory (EPROM), Programmable Read Only Memory (PROM), Read Only Memory (ROM), Magnetic Memory, Flash Memory, Disk or Optical Disk.
- SRAM static random access memory
- EEPROM electrically erasable programmable read only memory
- EPROM erasable Programmable Read Only Memory
- PROM Programmable Read Only Memory
- ROM Read Only Memory
- Magnetic Memory Flash Memory
- Disk Disk or Optical Disk.
- Power component 1306 provides power to various components of device 1300.
- Power component 1306 can include a power management system, one or more power sources, and other components associated with generating, managing, and distributing power for device 1300.
- the multimedia component 1313 includes a screen between the device 1300 and the user that provides an output interface.
- the screen can include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen can be implemented as a touch screen to receive input signals from the user.
- the touch panel includes one or more touch sensors to sense touches, slides, and gestures on the touch panel. The touch sensor may sense not only the boundary of the touch or sliding action, but also the duration and pressure associated with the touch or slide operation.
- the multimedia component 1313 includes a front camera and/or a rear camera. When the device 1300 is in an operation mode, such as a shooting mode or a video mode, the front camera and/or the rear camera can receive external multimedia data. Each front and rear camera can be a fixed optical lens system or have focal length and optical zoom capabilities.
- the audio component 1310 is configured to output and/or input an audio signal.
- the audio component 1310 includes a microphone (MIC) that, when the device 1300 is in an operational mode, such as a call mode, a recording mode, and a voice recognition mode, The microphone is configured to receive an external audio signal.
- the received audio signal may be further stored in memory 1304 or transmitted via communication component 1316.
- the audio component 1310 also includes a speaker for outputting an audio signal.
- the I/O interface 1312 provides an interface between the processing component 1302 and the peripheral interface module, which may be a keyboard, a click wheel, a button, or the like. These buttons may include, but are not limited to, a home button, a volume button, a start button, and a lock button.
- Sensor assembly 1314 includes one or more sensors for providing device 1300 with a status assessment of various aspects.
- sensor assembly 1314 can detect an open/closed state of device 1300, a relative positioning of components, such as the display and keypad of device 1300, and sensor component 1314 can also detect a change in position of one component of device 1300 or device 1300. The presence or absence of contact by the user with the device 1300, the orientation or acceleration/deceleration of the device 1300 and the temperature change of the device 1300.
- Sensor assembly 1314 can include a proximity sensor configured to detect the presence of nearby objects without any physical contact.
- Sensor assembly 1314 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications.
- the sensor component 1314 can also include an acceleration sensor, a gyro sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.
- Communication component 1316 is configured to facilitate wired or wireless communication between device 1300 and other devices.
- the device 1300 can access a wireless network based on a communication standard, such as WiFi, 2G or 3G, or a combination thereof.
- communication component 1316 receives broadcast signals or broadcast associated information from an external broadcast management system via a broadcast channel.
- the communication component 1316 also includes a near field communication (NFC) module to facilitate short range communication.
- NFC near field communication
- the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology, and other technologies.
- RFID radio frequency identification
- IrDA infrared data association
- UWB ultra-wideband
- Bluetooth Bluetooth
- apparatus 1300 may be implemented by one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable A gate array (FPGA), controller, microcontroller, microprocessor, or other electronic component implementation for performing the above methods.
- ASICs application specific integrated circuits
- DSPs digital signal processors
- DSPDs digital signal processing devices
- PLDs programmable logic devices
- FPGA field programmable A gate array
- controller microcontroller, microprocessor, or other electronic component implementation for performing the above methods.
- non-transitory computer readable storage medium comprising instructions, such as a memory 1304 comprising instructions executable by processor 1320 of apparatus 1300 to perform the above method.
- the non-transitory computer readable storage medium may be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, and an optical data storage device.
- a non-transitory computer readable storage medium when instructions in the storage medium are executed by a processor of a mobile terminal, enabling the mobile terminal to perform an image classification method, the method comprising:
- the face pose information in the two images to be classified are respectively determined according to the pre-established face pose classifier model
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Multimedia (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Software Systems (AREA)
- Probability & Statistics with Applications (AREA)
- Human Computer Interaction (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Medical Informatics (AREA)
- Image Analysis (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
一种图像分类方法及装置,应用于终端,其方法包括:当获取到都包含人脸特征信息的两个待分类图像时,根据预先建立的人脸姿态分类器模型,分别确定所述两个待分类图像中的人脸姿态信息(S110);根据所述两个待分类图像中的人脸姿态信息,确定所述两个待分类图像之间的目标聚类合并阈值(S120);利用确定得到的所述目标聚类合并阈值对所述两个待分类图像进行分类(S130)。通过确定将人脸姿态信息,可以更准确地确定两个待分类图像是否属于同一类图像,进而提高图像分类效率。
Description
本申请基于申请号为CN201510846109.1、申请日为2015年11月27日的中国专利申请提出,并要求该中国专利申请的优先权,该中国专利申请的全部内容在此引入本申请作为参考。
本公开涉及图像识别技术领域,尤其涉及一种图像分类方法及装置。
随着科技的不断发展,以智能手机为代表的智能终端可以实现越来越多的功能。以智能手机为例,根据需要用户可以利用智能手机进行自拍或给其他人拍照,也可以将自己或他人空间中的照片下载到本地进行存储。
然而,随着智能手机的存储空间的增大,用户在智能手机存储的照片也越来越多,对这些照片的管理也变得十分繁琐,很多情况下用户希望可以将同一个人的照片聚集在一起进行显示,以方便用户浏览。
发明内容
为克服相关技术中存在的问题,本公开提供一种图像分类方法及装置。
根据本公开实施例的第一方面,提供一种图像分类方法,包括:
当获取到都包含人脸特征信息的两个待分类图像时,根据预先建立的人脸姿态分类器模型,分别确定所述两个待分类图像中的人脸姿态信息;
根据所述两个待分类图像中的人脸姿态信息,确定所述两个待分类图像之间的目标聚类合并阈值;
利用确定得到的所述目标聚类合并阈值对所述两个待分类图像进行分类。
在对两个待分类图像进行分类时,根据预先建立的任亮姿态分类器模型,分别确定出这两个待分类图像中的人脸姿态信息,进而确定出这两个待分类图像之间的目标聚类合并阈值,然后根据该目标聚类合并阈值对这两个待分类图像进行分类。通过确定将人脸姿态信息,可以更准确地确定两个待分类图像是否属于同一类图像,进而提高图像分类效率。
可选地,还包括:
计算所述两个待分类图像之间的相似度;
判断所述相似度是否大于预设阈值;
当所述相似度大于预设阈值时,执行所述分别确定所述两个待分类图像中的人脸姿态信息的步骤。
针对两个具有一定相似度的图像进行分类,即在无法单单通过相似度区分二者是否为
同一类图像时,借助本公开提供的实施例可以进一步提高两个待分类图像的准确度。如果两个待分类图像小于某相似度,即在从确定两个待分类已经明显不属于同一类图像时,也就无需采用本公开提供的方法。因此,待分类的两个图像之间至少要满足具有一定的相似度才可以。
可选地,所述确定所述两个待分类图像之间的目标聚类合并阈值,包括:
获取预设聚类合并阈值;
判断所述两个待分类图像中的人脸姿态信息是否都是大角度姿态信息;
当所述两个待分类图像中的人脸姿态信息都是大角度姿态信息时,按预设方式增大所述预设聚类合并阈值,将所述增大后的所述预设聚类合并阈值作为目标聚类合并阈值。
如果两个待分类图像中的人脸姿态信息都是大角度姿态信息,说明这两个待分类图像中人脸特征含量丰富,需要调高二者之间的聚类合并阈值,以提高分类的准确度。
可选地,所述确定所述两个待分类图像之间的目标聚类合并阈值,包括:
获取预设聚类合并阈值;
判断所述两个待分类图像中的人脸姿态信息是否分别为不同角度姿态信息;
当所述两个待分类图像中的人脸姿态信息分别为不同角度姿态信息,按照预设方式减小所述预设聚类合并阈值,将减小后的所述预设聚类合并阈值作为目标聚类合并阈值。
如果两个待分类图像中的人脸姿态信息都是不同角度姿态信息,说明这两个待分类图像中人脸特征含量不太丰富,需要调低二者之间的聚类合并阈值,以提高分类的准确度。
可选地,所述确定所述两个待分类图像之间的目标聚类合并阈值,包括:
获取预设聚类合并阈值;
判断所述两个待分类图像中的人脸姿态信息是否都为正面姿态信息;
当所述两个待分类图像中的人脸姿态信息都为正面姿态信息时,将所述预设聚类合并阈值作为目标聚类合并阈值。
如果两个待分类图像中的人脸姿态信息都是正面姿态信息,说明这两个待分类图像中包含全部人脸特征,这是可以直接将预设聚类合并阈值作为目标聚类合并阈值。
可选地,所述利用确定得到的所述目标聚类合并阈值对所述两个待分类图像进行分类,包括:
计算所述两个待分类图像的聚类合并值;
判断所述聚类合并值是否大于所述目标聚类合并阈值;
当所述聚类合并值大于所述目标聚类合并阈值时,将所述两个待分类图像作为同一类图像;
当所述聚类合并值不大于所述目标聚类合并阈值时,将所述两个待分类图像作为不同类图像。
通过将两个待分类图像的聚类合并值与目标聚类合并阈值进行比较,可以很方便准确的判断出这两个待分类图像是否属于同一类图像。
根据本公开实施例的第二方面,提供一种图像分类装置,包括:
人脸姿态信息确定模块,用于在获取到都包含人脸特征信息的两个待分类图像时,根据预先建立的人脸姿态分类器模型,分别确定所述两个待分类图像中的人脸姿态信息;
目标聚类合并阈值确定模块,用于根据所述两个待分类图像中的人脸姿态信息,确定所述两个待分类图像之间的目标聚类合并阈值;
图像分类模块,用于利用确定得到的所述目标聚类合并阈值对所述两个待分类图像进行分类。
可选地,还包括:
相似度计算模块,用于计算所述两个待分类图像之间的相似度;
阈值判断模块,用于判断所述相似度是否大于预设阈值。
可选地,所述目标聚类合并阈值确定模块,包括:
第一聚类合并阈值获取子模块,用于获取预设聚类合并阈值;
大角度姿态信息判断子模块,用于判断所述两个待分类图像中的人脸姿态信息是否都是大角度姿态信息;
第一目标聚类合并阈值子模块,用于当所述两个待分类图像中的人脸姿态信息都是大角度姿态信息时,按预设方式增大所述预设聚类合并阈值,将所述增大后的所述预设聚类合并阈值作为目标聚类合并阈值。
可选地,所述目标聚类合并阈值确定模块,包括:
第二聚类合并阈值获取子模块,用于获取预设聚类合并阈值;
不同角度姿态信息判断子模块,用于判断所述两个待分类图像中的人脸姿态信息是否分别为不同角度姿态信息;
第二目标聚类合并阈值子模块,用于在所述两个待分类图像中的人脸姿态信息分别为不同角度姿态信息,按照预设方式减小所述预设聚类合并阈值,将减小后的所述预设聚类合并阈值作为目标聚类合并阈值。
可选地,所述目标聚类合并阈值确定模块,包括:
第三聚类合并阈值获取子模块,用于获取预设聚类合并阈值;
正面姿态信息判断子模块,用于判断所述两个待分类图像中的人脸姿态信息是否都为正面姿态信息;
第三目标聚类合并阈值子模块,用于在所述两个待分类图像中的人脸姿态信息都为正面姿态信息时,将所述预设聚类合并阈值作为目标聚类合并阈值。
可选地,所述图像分类模块,包括:
聚类合并值计算子模块,用于计算所述两个待分类图像的聚类合并值;
聚类合并值判断子模块,用于判断所述聚类合并值是否大于所述目标聚类合并阈值;
同一类图像确定子模块,用于在所述聚类合并值大于所述目标聚类合并阈值时,将所述两个待分类图像作为同一类图像;
不同类图像确定子模块,用于在所述聚类合并值不大于所述目标聚类合并阈值时,将所述两个待分类图像作为不同类图像。
根据本公开实施例的第三方面,提供一种终端,包括:
处理器;
用于存储处理器可执行指令的存储器;
其中,所述处理器被配置为:
当获取到都包含人脸特征信息的两个待分类图像时,根据预先建立的人脸姿态分类器模型,分别确定所述两个待分类图像中的人脸姿态信息;
根据所述两个待分类图像中的人脸姿态信息,确定所述两个待分类图像之间的目标聚类合并阈值;
利用确定得到的所述目标聚类合并阈值对所述两个待分类图像进行分类。
本公开的实施例提供的技术方案可以包括以下有益效果:
本公开实施例中提供的图像分类方法及装置,在对两个待分类图像进行分类时,根据预先建立的任亮姿态分类器模型,分别确定出这两个待分类图像中的人脸姿态信息,进而确定出这两个待分类图像之间的目标聚类合并阈值,然后根据该目标聚类合并阈值对这两个待分类图像进行分类。通过确定将人脸姿态信息,可以更准确地确定两个待分类图像是否属于同一类图像,进而提高图像分类效率。
应当理解的是,以上的一般描述和后文的细节描述仅是示例性和解释性的,并不能限制本公开。
此处的附图被并入说明书中并构成本说明书的一部分,示出了符合本发明的实施例,并与说明书一起用于解释本发明的原理。
图1是根据一示例性实施例示出的一种图像分类方法的流程图;
图2是根据另一示例性实施例示出的一种图像分类方法的流程图;
图3是图1中步骤S120的流程图;
图4是图1中步骤S120的另一流程图;
图5是图1中步骤S120的又一流程图;
图6是图1中步骤S130的流程图;
图7是根据一示例性实施例示出的一种图像分类装置的示意图;
图8是根据又一示例性实施例示出的一种图像分类装置的示意图;
图9是图7中目标聚类合并阈值确定模块的示意图;
图10是图7中目标聚类合并阈值确定模块的另一示意图;
图11是图7中目标聚类合并阈值确定模块的又一示意图;
图12是图7中图像分类模块的示意图;
图13是根据一示例性实施例示出的一种终端的框图。
这里将详细地对示例性实施例进行说明,其示例表示在附图中。下面的描述涉及附图时,除非另有表示,不同附图中的相同数字表示相同或相似的要素。以下示例性实施例中所描述的实施方式并不代表与本发明相一致的所有实施方式。相反,它们仅是与如所附权利要求书中所详述的、本发明的一些方面相一致的装置和方法的例子。
为了解决相关技术问题。本公开实施例首先提供了一种图像分类方法,如图1所示,可以包括如下步骤:
在步骤S110中,当获取到都包含人脸特征信息的两个待分类图像时,根据预先建立的人脸姿态分类器模型,分别确定两个待分类图像中的人脸姿态信息。
在本公开实施例中,可以通过预先建立的人脸姿态分类器模型,确定出两个待分类图像中的人脸姿态信息。如:正脸、侧脸,左脸、有脸,或者两个正脸图像等。
在步骤S120中,根据两个待分类图像中的人脸姿态信息,确定两个待分类图像之间的目标聚类合并阈值。
本公开实施例中,在根据通过预先建立的人脸姿态分类器模型,分别确定出两个待分类图像的人脸姿态信息之后,还需要根据该人脸姿态信息,获取与其相对应的聚类合并阈值。另外,在对包括多张人脸图像照片进行分类时,首先可以对其中的任两个图像照片进行识别,判定是否为同一个人的照片,然后得到多张人脸图像照片的最终分类结果。还可以首先任取一张图像照片,分别对其他图像照片进行逐一对比,得到这一张图像照片对其他图像照片的分类结果,然后用同样的方式确定出剩下的其他图像照片的分类结果。
一张图像照片可以是,正面人脸姿态信息或其他角度的侧面人脸姿态信息。例如,两张图像照片可以是,两张正面人脸图像照片、一张正面人脸和一张侧面人脸、两张侧面人脸等情况,每种情况都会对应一个聚类合并阈值,其中侧面人脸姿态信息可以包括不同角度的侧面人脸姿态信息。根据两个待分类图像中的人脸姿态信息,可以通过查表或自动生成等方式确定出两个待分类图像之间的聚类合并阈值作为目标聚类合并阈值。
在步骤S130中,利用确定得到的目标聚类合并阈值对两个待分类图像进行分类。
由于本公开实施例是对两个相似度大于设定阈值的待分类图像进行再次的判别,进一步提高了分类的准确性,所以两个待分类图像的聚类合值大于目标聚类合并阈值时,说明这两个待分类图像属于一类,否则不属于一类。
本公开实施例中提供的图像分类方法,在对两个待分类图像进行分类时,根据预先建立的任亮姿态分类器模型,分别确定出这两个待分类图像中的人脸姿态信息,进而确定出这两个待分类图像之间的目标聚类合并阈值,然后根据该目标聚类合并阈值对这两个待分类图像进行分类。通过确定将人脸姿态信息,可以更准确地确定两个待分类图像是否属于同一类图像,进而提高图像分类效率。
作为对图1方法的细化,在另一实施例中,如图2所示,该方法还可以包括以下步骤:
在步骤S140中,计算两个待分类图像之间的相似度。
在步骤S150中,判断相似度是否大于预设阈值。
当相似度大于预设阈值时,执行步骤S110。
当相似度不大于预设阈值时,结束本次流程。
本公开实施例主要是对两个具有一定相似度的图像进行分类,即在无法单单通过相似度区分二者是否为同一类图像时,借助本公开提供的实施例可以进一步提高两个待分类图像的准确度。如果两个待分类图像小于某相似度,即在从确定两个待分类已经明显不属于同一类图像时,也就无需采用本公开提供的方法。因此,待分类的两个图像之间至少要满足具有一定的相似度才可以。
为了确定两个待分类图像中的人脸姿态信息都是大角度姿态信息时的聚类合并阈值,进而根据该聚类合并阈值进行分类,作为对图1方法的细化,在另一实施例中,如图3所示,步骤S120还可以包括以下步骤:
在步骤S121中,获取预设聚类合并阈值。
在步骤S122中,判断两个待分类图像中的人脸姿态信息是否都是大角度姿态信息。
当两个待分类图像中的人脸姿态信息都是大角度姿态信息时,在步骤S123中,按预设方式增大预设聚类合并阈值,将增大后的预设聚类合并阈值作为目标聚类合并阈值。
如果两个待分类图像中的人脸姿态信息都是大角度姿态信息,说明这两个待分类图像中人脸特征含量丰富,需要调高二者之间的聚类合并阈值,以提高分类的准确度。
为了确定两个待分类图像中的人脸姿态信息分别为不同角度姿态信息时的聚类合并阈值,进而根据该聚类合并阈值进行分类,作为对图1方法的细化,在另一实施例中,如图4所示,步骤S120还可以包括以下步骤:
在步骤S124中,获取预设聚类合并阈值。
在步骤S125中,判断两个待分类图像中的人脸姿态信息是否分别为不同角度姿态信息。
当两个待分类图像中的人脸姿态信息分别为不同角度姿态信息,在步骤S126中,按照预设方式减小预设聚类合并阈值,将减小后的预设聚类合并阈值作为目标聚类合并阈值。
如果两个待分类图像中的人脸姿态信息都是不同角度姿态信息,说明这两个待分类图像中人脸特征含量不太丰富,需要调低二者之间的聚类合并阈值,以提高分类的准确度。
为了确定两个待分类图像中的人脸姿态信息都为正面姿态信息时的聚类合并阈值,进而根据该聚类合并阈值进行分类,作为对图1方法的细化,在另一实施例中,如图5所示,步骤S120还可以包括以下步骤:
在步骤S127中,获取预设聚类合并阈值。
在步骤S128中,判断两个待分类图像中的人脸姿态信息是否都为正面姿态信息。
当两个待分类图像中的人脸姿态信息都为正面姿态信息时,在步骤S129中,将预设聚类合并阈值作为目标聚类合并阈值。
如果两个待分类图像中的人脸姿态信息都是正面姿态信息,说明这两个待分类图像中包含全部人脸特征,这是可以直接将预设聚类合并阈值作为目标聚类合并阈值。
下面着重对上述三种人脸姿态信息进行详细说明,即:都是大角度姿态信息、不同角度姿态信息和都为正面姿态信息。
为了判别待分类图像中的人脸姿态信息,如待分类图像中的人脸是正面的、还是侧面的以一定角度呈现出来的,需要提取出待分类图像中的人脸特征信息,通过该人脸特征信息进行判别该待分类图像中的人脸姿态信息。例如,可以对正面人脸建立水平、竖直和垂直方向上的三维坐标系,提取图像中的人眼、鼻子、耳朵等方位等特征信息来判断该待分类图像的人脸姿态。
提取待分类图像中的人脸特征信息,通过预先建立的人脸姿态分类器模型,判断出该待分类图像中的人脸姿态信息,如该图像中的人脸是朝某个位的偏移量,如该待分类图像中的人脸姿态信息为左偏20度。
具体可以通过采集多种人脸姿态样本图像照片,建立人脸姿态样本库。其中,该人脸姿态样本库可以包括多张多种人脸姿态的照片图像。如:该人脸姿态样本库包括:正面人脸姿态照片图像1000张、10度人脸姿态照片图像1000张、20度人脸姿态照片图像1000张等等。示例性的,10度人脸姿态照片可以是拍摄时,人脸正面与镜头之间的夹角。其中,不同角度和照片的数量可以根据实际的需要进行设定,为了提高人脸姿态分类器模型对人脸姿态判别的准确性,人脸姿态样本库中包含的人脸照片应当包括男人和女人的照片图像,又如可以包括老人和孩子的照片图像等等。
在上述人脸姿态样本库建好之后,可以通过现有的分类器对建立的人脸姿态样本库中的照片图像进行训练,得到人脸姿态分类器模型。例如,可以采用非线性分类器对人脸姿态样本库中的图像照片进行训练人脸姿态分类器模型。示例性的,可以采用SVM(Support Vector Machine,支持向量机)或CNN(convolutional neural networks,卷积神经网络)对人脸姿态样本库中的照片图像进行训练,得到人脸姿态分类器模型。
作为对图1方法的细化,在另一实施例中,如图6所示,步骤S130还可以包括以下步骤:
在步骤S131中,计算两个待分类图像的聚类合并值。
在步骤S132中,判断聚类合并值是否大于目标聚类合并阈值。
当聚类合并值大于目标聚类合并阈值时,在步骤S133中,将两个待分类图像作为同一类图像。
当聚类合并值不大于目标聚类合并阈值时,在步骤S134中,将两个待分类图像作为不同类图像。
通过将两个待分类图像的聚类合并值与目标聚类合并阈值进行比较,可以很方便准确
的判断出这两个待分类图像是否属于同一类图像。
本公开实施例中提供的图像分类方法,在对两个待分类图像进行分类时,根据预先建立的任亮姿态分类器模型,分别确定出这两个待分类图像中的人脸姿态信息,进而确定出这两个待分类图像之间的目标聚类合并阈值,然后根据该目标聚类合并阈值对这两个待分类图像进行分类。通过确定将人脸姿态信息,可以更准确地确定两个待分类图像是否属于同一类图像,进而提高图像分类效率。
通过以上的方法实施例的描述,所属领域的技术人员可以清楚地了解到本公开可借助软件加必需的通用硬件平台的方式来实现,当然也可以通过硬件,但很多情况下前者是更佳的实施方式。基于这样的理解,本公开的技术方案本质上或者说对现有技术做出贡献的部分可以以软件产品的形式体现出来,该计算机软件产品存储在一个存储介质中,包括若干指令用以使得一台计算机设备(可以是个人计算机,服务器,或者网络设备等)执行本公开各个实施例所述方法的全部或部分步骤。而前述的存储介质包括:只读存储器(ROM)、随机存取存储器(RAM)、磁碟或者光盘等各种可以存储程序代码的介质。
另外,作为对上述各实施例的实现,本公开实施例还提供了一种图像分类装置,该装置位于终端中,如图7所示,该装置包括:人脸姿态信息确定模块10、目标聚类合并阈值确定模块20和图像分类模块30,其中,
人脸姿态信息确定模块10被配置为在获取到都包含人脸特征信息的两个待分类图像时,根据预先建立的人脸姿态分类器模型,分别确定所述两个待分类图像中的人脸姿态信息;
在本公开实施例中,可以通过预先建立的人脸姿态分类器模型,确定出两个待分类图像中的人脸姿态信息。如:正脸、侧脸,左脸、有脸,或者两个正脸图像等。
目标聚类合并阈值确定模块20被配置为根据所述两个待分类图像中的人脸姿态信息,确定所述两个待分类图像之间的目标聚类合并阈值;
本公开实施例中,在根据通过预先建立的人脸姿态分类器模型,分别确定出两个待分类图像的人脸姿态信息之后,还需要根据该人脸姿态信息,获取与其相对应的聚类合并阈值。另外,在对包括多张人脸图像照片进行分类时,首先可以对其中的任两个图像照片进行识别,判定是否为同一个人的照片,然后得到多张人脸图像照片的最终分类结果。还可以首先任取一张图像照片,分别对其他图像照片进行逐一对比,得到这一张图像照片对其他图像照片的分类结果,然后用同样的方式确定出剩下的其他图像照片的分类结果。
一张图像照片可以是,正面人脸姿态信息或其他角度的侧面人脸姿态信息。例如,两张图像照片可以是,两张正面人脸图像照片、一张正面人脸和一张侧面人脸、两张侧面人脸等情况,每种情况都会对应一个聚类合并阈值,其中侧面人脸姿态信息可以包括不同角度的侧面人脸姿态信息。根据两个待分类图像中的人脸姿态信息,可以通过查表或自动生成等方式确定出两个待分类图像之间的聚类合并阈值作为目标聚类合并阈值。
图像分类模块30被配置为利用确定得到的所述目标聚类合并阈值对所述两个待分类
图像进行分类。
由于本公开实施例是对两个相似度大于设定阈值的待分类图像进行再次的判别,进一步提高了分类的准确性,所以两个待分类图像的聚类合值大于目标聚类合并阈值时,说明这两个待分类图像属于一类,否则不属于一类。
本公开实施例中提供的图像分类装置,在对两个待分类图像进行分类时,根据预先建立的任亮姿态分类器模型,分别确定出这两个待分类图像中的人脸姿态信息,进而确定出这两个待分类图像之间的目标聚类合并阈值,然后根据该目标聚类合并阈值对这两个待分类图像进行分类。通过确定将人脸姿态信息,可以更准确地确定两个待分类图像是否属于同一类图像,进而提高图像分类效率。
在本公开提供的又一实施例中,基于图7,如图8所示,该装置还包括:相似度计算模块40和阈值判断模块50,其中,
相似度计算模块40被配置为计算所述两个待分类图像之间的相似度;
阈值判断模块50被配置为判断所述相似度是否大于预设阈值。
本公开实施例主要是对两个具有一定相似度的图像进行分类,即在无法单单通过相似度区分二者是否为同一类图像时,借助本公开提供的实施例可以进一步提高两个待分类图像的准确度。如果两个待分类图像小于某相似度,即在从确定两个待分类已经明显不属于同一类图像时,也就无需采用本公开提供的方法。因此,待分类的两个图像之间至少要满足具有一定的相似度才可以。
在本公开提供的又一实施例中,基于图7,如图9所示,目标聚类合并阈值确定模块20,包括:第一聚类合并阈值获取子模块21、大角度姿态信息判断子模块22和第一目标聚类合并阈值子模块23,其中,
第一聚类合并阈值获取子模块21被配置为获取预设聚类合并阈值;
大角度姿态信息判断子模块22被配置为判断所述两个待分类图像中的人脸姿态信息是否都是大角度姿态信息;
第一目标聚类合并阈值子模块23被配置为当所述两个待分类图像中的人脸姿态信息都是大角度姿态信息时,按预设方式增大所述预设聚类合并阈值,将所述增大后的所述预设聚类合并阈值作为目标聚类合并阈值。
如果两个待分类图像中的人脸姿态信息都是大角度姿态信息,说明这两个待分类图像中人脸特征含量丰富,需要调高二者之间的聚类合并阈值,以提高分类的准确度。
在本公开提供的又一实施例中,基于图7,如图10所示,所述目标聚类合并阈值确定模块20,包括:第二聚类合并阈值获取子模块24、不同角度姿态信息判断子模块25和第二目标聚类合并阈值子模块26,其中,
第二聚类合并阈值获取子模块24被配置为获取预设聚类合并阈值;
不同角度姿态信息判断子模块25被配置为判断所述两个待分类图像中的人脸姿态信息是否分别为不同角度姿态信息;
第二目标聚类合并阈值子模块26被配置为在所述两个待分类图像中的人脸姿态信息分别为不同角度姿态信息,按照预设方式减小所述预设聚类合并阈值,将减小后的所述预设聚类合并阈值作为目标聚类合并阈值。
如果两个待分类图像中的人脸姿态信息都是不同角度姿态信息,说明这两个待分类图像中人脸特征含量不太丰富,需要调低二者之间的聚类合并阈值,以提高分类的准确度。
在本公开提供的又一实施例中,基于图7,如图11所示,所述目标聚类合并阈值确定模块20,包括:第三聚类合并阈值获取子模块27、正面姿态信息判断子模块28和第三目标聚类合并阈值子模块29,其中,
第三聚类合并阈值获取子模块27被配置为获取预设聚类合并阈值;
正面姿态信息判断子模块28被配置为判断所述两个待分类图像中的人脸姿态信息是否都为正面姿态信息;
第三目标聚类合并阈值子模块29被配置为在所述两个待分类图像中的人脸姿态信息都为正面姿态信息时,将所述预设聚类合并阈值作为目标聚类合并阈值。
如果两个待分类图像中的人脸姿态信息都是正面姿态信息,说明这两个待分类图像中包含全部人脸特征,这是可以直接将预设聚类合并阈值作为目标聚类合并阈值。
在本公开提供的又一实施例中,基于图7,如图12所示,所述图像分类模块30,包括:聚类合并值计算子模块31、聚类合并值判断子模块32、同一类图像确定子模块33和不同类图像确定子模块34,其中,
聚类合并值计算子模块31被配置为计算所述两个待分类图像的聚类合并值;
聚类合并值判断子模块32被配置为判断所述聚类合并值是否大于所述目标聚类合并阈值;
同一类图像确定子模块33被配置为在所述聚类合并值大于所述目标聚类合并阈值时,将所述两个待分类图像作为同一类图像;
不同类图像确定子模块34被配置为在所述聚类合并值不大于所述目标聚类合并阈值时,将所述两个待分类图像作为不同类图像。
通过将两个待分类图像的聚类合并值与目标聚类合并阈值进行比较,可以很方便准确的判断出这两个待分类图像是否属于同一类图像。
本公开实施例中提供的图像分类装置,在对两个相似度较大的待分类图像进行分类时,获得这两个待分类图像人脸的姿态信息,通过判断这两个相似度较大的待分类图像的姿态信息,确定出与这两个待分类图像的姿态信息相对应的聚类合并阈值。当这两个待分类图像相似度大于该聚类合并阈值时,将这两个待分类图像分为一类。可以有效避免只通过图像中的相似度直接判断两个待分类图像是否为一类图像,进而造成对图像分类的错误率较高的问题。
另外,根据当两个待分类图像中包含的人脸姿态信息不同,两个待分类图像的姿态信息相对应的聚类合并阈值是可变的。如果这两个待分类图像的姿态信息比较接近,例如都
是大角度姿态时,可以将聚类合并阈值调的高一点;如果这两个待分类图像的姿态信息差异较大,如一个是左侧人脸姿态信息,另一个是右侧人脸姿态信息,这时可以将二者之间的聚类合并阈值调的低一点;如果这两个待分类图像的姿态信息都属于正面人脸姿态信息,可以不对二者之间的聚类合并阈值调进行调整。即可以使两个待分类图像的姿态信息相对应的聚类合并阈值具有自适应性,可以准确的对两个待分类图像进行准确判定是否为同一个人的照片图像。
图13是根据一示例性实施例示出的一种用于图像分类的装置1300的结构示意图。例如,装置1300可以是移动电话,计算机,数字广播终端,消息收发设备,游戏控制台,平板设备,医疗设备,健身设备,个人数字助理等。
参照图13,装置1300可以包括以下一个或多个组件:处理组件1302,存储器1304,电源组件1306,多媒体组件1313,音频组件1310,输入/输出(I/O)的接口1312,传感器组件1314,以及通信组件1316。
处理组件1302通常控制装置1300的整体操作,诸如与显示,电话呼叫,数据通信,相机操作和记录操作相关联的操作。处理组件1302可以包括一个或多个处理器1320来执行指令,以完成上述的方法的全部或部分步骤。此外,处理组件1302可以包括一个或多个模块,便于处理组件1302和其他组件之间的交互。例如,处理组件1302可以包括多媒体模块,以方便多媒体组件1313和处理组件1302之间的交互。
存储器1304被配置为存储各种类型的数据以支持在装置1300的操作。这些数据的示例包括用于在装置1300上操作的任何应用程序或方法的指令,联系人数据,电话簿数据,消息,图片,视频等。存储器1304可以由任何类型的易失性或非易失性存储设备或者它们的组合实现,如静态随机存取存储器(SRAM),电可擦除可编程只读存储器(EEPROM),可擦除可编程只读存储器(EPROM),可编程只读存储器(PROM),只读存储器(ROM),磁存储器,快闪存储器,磁盘或光盘。
电源组件1306为装置1300的各种组件提供电力。电源组件1306可以包括电源管理系统,一个或多个电源,及其他与为装置1300生成、管理和分配电力相关联的组件。
多媒体组件1313包括在所述装置1300和用户之间的提供一个输出接口的屏幕。在一些实施例中,屏幕可以包括液晶显示器(LCD)和触摸面板(TP)。如果屏幕包括触摸面板,屏幕可以被实现为触摸屏,以接收来自用户的输入信号。触摸面板包括一个或多个触摸传感器以感测触摸、滑动和触摸面板上的手势。所述触摸传感器可以不仅感测触摸或滑动动作的边界,而且还检测与所述触摸或滑动操作相关的持续时间和压力。在一些实施例中,多媒体组件1313包括一个前置摄像头和/或后置摄像头。当装置1300处于操作模式,如拍摄模式或视频模式时,前置摄像头和/或后置摄像头可以接收外部的多媒体数据。每个前置摄像头和后置摄像头可以是一个固定的光学透镜系统或具有焦距和光学变焦能力。
音频组件1310被配置为输出和/或输入音频信号。例如,音频组件1310包括一个麦克风(MIC),当装置1300处于操作模式,如呼叫模式、记录模式和语音识别模式时,
麦克风被配置为接收外部音频信号。所接收的音频信号可以被进一步存储在存储器1304或经由通信组件1316发送。在一些实施例中,音频组件1310还包括一个扬声器,用于输出音频信号。
I/O接口1312为处理组件1302和外围接口模块之间提供接口,上述外围接口模块可以是键盘,点击轮,按钮等。这些按钮可包括但不限于:主页按钮、音量按钮、启动按钮和锁定按钮。
传感器组件1314包括一个或多个传感器,用于为装置1300提供各个方面的状态评估。例如,传感器组件1314可以检测到装置1300的打开/关闭状态,组件的相对定位,例如所述组件为装置1300的显示器和小键盘,传感器组件1314还可以检测装置1300或装置1300一个组件的位置改变,用户与装置1300接触的存在或不存在,装置1300方位或加速/减速和装置1300的温度变化。传感器组件1314可以包括接近传感器,被配置用来在没有任何的物理接触时检测附近物体的存在。传感器组件1314还可以包括光传感器,如CMOS或CCD图像传感器,用于在成像应用中使用。在一些实施例中,该传感器组件1314还可以包括加速度传感器,陀螺仪传感器,磁传感器,压力传感器或温度传感器。
通信组件1316被配置为便于装置1300和其他设备之间有线或无线方式的通信。装置1300可以接入基于通信标准的无线网络,如WiFi,2G或3G,或它们的组合。在一个示例性实施例中,通信组件1316经由广播信道接收来自外部广播管理系统的广播信号或广播相关信息。在一个示例性实施例中,所述通信组件1316还包括近场通信(NFC)模块,以促进短程通信。例如,在NFC模块可基于射频识别(RFID)技术,红外数据协会(IrDA)技术,超宽带(UWB)技术,蓝牙(BT)技术和其他技术来实现。
在示例性实施例中,装置1300可以被一个或多个应用专用集成电路(ASIC)、数字信号处理器(DSP)、数字信号处理设备(DSPD)、可编程逻辑器件(PLD)、现场可编程门阵列(FPGA)、控制器、微控制器、微处理器或其他电子元件实现,用于执行上述方法。
在示例性实施例中,还提供了一种包括指令的非临时性计算机可读存储介质,例如包括指令的存储器1304,上述指令可由装置1300的处理器1320执行以完成上述方法。例如,所述非临时性计算机可读存储介质可以是ROM、随机存取存储器(RAM)、CD-ROM、磁带、软盘和光数据存储设备等。
一种非临时性计算机可读存储介质,当所述存储介质中的指令由移动终端的处理器执行时,使得移动终端能够执行一种图像分类方法,所述方法包括:
当获取到都包含人脸特征信息的两个待分类图像时,根据预先建立的人脸姿态分类器模型,分别确定所述两个待分类图像中的人脸姿态信息;
根据所述两个待分类图像中的人脸姿态信息,确定所述两个待分类图像之间的目标聚类合并阈值;
利用确定得到的所述目标聚类合并阈值对所述两个待分类图像进行分类。
本领域技术人员在考虑说明书及实践这里公开的发明后,将容易想到本发明的其它实施方案。本申请旨在涵盖本发明的任何变型、用途或者适应性变化,这些变型、用途或者适应性变化遵循本发明的一般性原理并包括本公开未公开的本技术领域中的公知常识或惯用技术手段。说明书和实施例仅被视为示例性的,本发明的真正范围和精神由下面的权利要求指出。
应当理解的是,本发明并不局限于上面已经描述并在附图中示出的精确结构,并且可以在不脱离其范围进行各种修改和改变。本发明的范围仅由所附的权利要求来限制。
Claims (13)
- 一种图像分类方法,其特征在于,包括:当获取到都包含人脸特征信息的两个待分类图像时,根据预先建立的人脸姿态分类器模型,分别确定所述两个待分类图像中的人脸姿态信息;根据所述两个待分类图像中的人脸姿态信息,确定所述两个待分类图像之间的目标聚类合并阈值;利用确定得到的所述目标聚类合并阈值对所述两个待分类图像进行分类。
- 根据权利要求1所述的方法,其特征在于,还包括:计算所述两个待分类图像之间的相似度;判断所述相似度是否大于预设阈值;当所述相似度大于预设阈值时,执行所述分别确定所述两个待分类图像中的人脸姿态信息的步骤。
- 根据权利要求1或2所述的方法,其特征在于,所述确定所述两个待分类图像之间的目标聚类合并阈值,包括:获取预设聚类合并阈值;判断所述两个待分类图像中的人脸姿态信息是否都是大角度姿态信息;当所述两个待分类图像中的人脸姿态信息都是大角度姿态信息时,按预设方式增大所述预设聚类合并阈值,将所述增大后的所述预设聚类合并阈值作为目标聚类合并阈值。
- 根据权利要求1或2所述的方法,其特征在于,所述确定所述两个待分类图像之间的目标聚类合并阈值,包括:获取预设聚类合并阈值;判断所述两个待分类图像中的人脸姿态信息是否分别为不同角度姿态信息;当所述两个待分类图像中的人脸姿态信息分别为不同角度姿态信息,按照预设方式减小所述预设聚类合并阈值,将减小后的所述预设聚类合并阈值作为目标聚类合并阈值。
- 根据权利要求1或2所述的方法,其特征在于,所述确定所述两个待分类图像之间的目标聚类合并阈值,包括:获取预设聚类合并阈值;判断所述两个待分类图像中的人脸姿态信息是否都为正面姿态信息;当所述两个待分类图像中的人脸姿态信息都为正面姿态信息时,将所述预设聚类合并阈值作为目标聚类合并阈值。
- 根据权利要求1所述的方法,其特征在于,所述利用确定得到的所述目标聚类合并阈值对所述两个待分类图像进行分类,包括:计算所述两个待分类图像的聚类合并值;判断所述聚类合并值是否大于所述目标聚类合并阈值;当所述聚类合并值大于所述目标聚类合并阈值时,将所述两个待分类图像作为同一类 图像;当所述聚类合并值不大于所述目标聚类合并阈值时,将所述两个待分类图像作为不同类图像。
- 一种图像分类装置,其特征在于,包括:人脸姿态信息确定模块,用于在获取到都包含人脸特征信息的两个待分类图像时,根据预先建立的人脸姿态分类器模型,分别确定所述两个待分类图像中的人脸姿态信息;目标聚类合并阈值确定模块,用于根据所述两个待分类图像中的人脸姿态信息,确定所述两个待分类图像之间的目标聚类合并阈值;图像分类模块,用于利用确定得到的所述目标聚类合并阈值对所述两个待分类图像进行分类。
- 根据权利要求7所述的装置,其特征在于,还包括:相似度计算模块,用于计算所述两个待分类图像之间的相似度;阈值判断模块,用于判断所述相似度是否大于预设阈值。
- 根据权利要求7或8所述的装置,其特征在于,所述目标聚类合并阈值确定模块,包括:第一聚类合并阈值获取子模块,用于获取预设聚类合并阈值;大角度姿态信息判断子模块,用于判断所述两个待分类图像中的人脸姿态信息是否都是大角度姿态信息;第一目标聚类合并阈值子模块,用于当所述两个待分类图像中的人脸姿态信息都是大角度姿态信息时,按预设方式增大所述预设聚类合并阈值,将所述增大后的所述预设聚类合并阈值作为目标聚类合并阈值。
- 根据权利要求7或8所述的装置,其特征在于,所述目标聚类合并阈值确定模块,包括:第二聚类合并阈值获取子模块,用于获取预设聚类合并阈值;不同角度姿态信息判断子模块,用于判断所述两个待分类图像中的人脸姿态信息是否分别为不同角度姿态信息;第二目标聚类合并阈值子模块,用于在所述两个待分类图像中的人脸姿态信息分别为不同角度姿态信息,按照预设方式减小所述预设聚类合并阈值,将减小后的所述预设聚类合并阈值作为目标聚类合并阈值。
- 根据权利要求7或8所述的装置,其特征在于,所述目标聚类合并阈值确定模块,包括:第三聚类合并阈值获取子模块,用于获取预设聚类合并阈值;正面姿态信息判断子模块,用于判断所述两个待分类图像中的人脸姿态信息是否都为正面姿态信息;第三目标聚类合并阈值子模块,用于在所述两个待分类图像中的人脸姿态信息都为正面姿态信息时,将所述预设聚类合并阈值作为目标聚类合并阈值。
- 根据权利要求7所述的装置,其特征在于,所述图像分类模块,包括:聚类合并值计算子模块,用于计算所述两个待分类图像的聚类合并值;聚类合并值判断子模块,用于判断所述聚类合并值是否大于所述目标聚类合并阈值;同一类图像确定子模块,用于在所述聚类合并值大于所述目标聚类合并阈值时,将所述两个待分类图像作为同一类图像;不同类图像确定子模块,用于在所述聚类合并值不大于所述目标聚类合并阈值时,将所述两个待分类图像作为不同类图像。
- 一种终端,其特征在于,包括:处理器;用于存储处理器可执行指令的存储器;其中,所述处理器被配置为:当获取到都包含人脸特征信息的两个待分类图像时,根据预先建立的人脸姿态分类器模型,分别确定所述两个待分类图像中的人脸姿态信息;根据所述两个待分类图像中的人脸姿态信息,确定所述两个待分类图像之间的目标聚类合并阈值;利用确定得到的所述目标聚类合并阈值对所述两个待分类图像进行分类。
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510846109.1A CN105488527B (zh) | 2015-11-27 | 2015-11-27 | 图像分类方法及装置 |
CN201510846109.1 | 2015-11-27 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2017088470A1 true WO2017088470A1 (zh) | 2017-06-01 |
Family
ID=55675498
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2016/087562 WO2017088470A1 (zh) | 2015-11-27 | 2016-06-29 | 图像分类方法及装置 |
Country Status (4)
Country | Link |
---|---|
US (1) | US10282597B2 (zh) |
EP (1) | EP3176727A1 (zh) |
CN (1) | CN105488527B (zh) |
WO (1) | WO2017088470A1 (zh) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110929679A (zh) * | 2019-12-05 | 2020-03-27 | 杭州电子科技大学 | 一种基于gan的无监督自适应行人重识别方法 |
CN111081317A (zh) * | 2019-12-10 | 2020-04-28 | 山东大学 | 一种基于基因谱的乳腺癌淋巴结转移预测方法及预测系统 |
CN112818867A (zh) * | 2021-02-02 | 2021-05-18 | 浙江大华技术股份有限公司 | 人像聚类方法、设备及存储介质 |
Families Citing this family (20)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105488527B (zh) * | 2015-11-27 | 2020-01-10 | 小米科技有限责任公司 | 图像分类方法及装置 |
CN108229674B (zh) * | 2017-02-21 | 2021-03-02 | 北京市商汤科技开发有限公司 | 聚类用神经网络的训练方法和装置、聚类方法和装置 |
CN107909104B (zh) * | 2017-11-13 | 2023-07-18 | 腾讯数码(天津)有限公司 | 一种图片的人脸聚类方法、装置及存储介质 |
CN108875522B (zh) * | 2017-12-21 | 2022-06-10 | 北京旷视科技有限公司 | 人脸聚类方法、装置和系统及存储介质 |
CN108280190A (zh) * | 2018-01-24 | 2018-07-13 | 深圳前海大数金融服务有限公司 | 图像分类方法、服务器及存储介质 |
CN108960145A (zh) * | 2018-07-04 | 2018-12-07 | 北京蜂盒科技有限公司 | 人脸图像检测方法、装置、存储介质和电子设备 |
CN110942081B (zh) * | 2018-09-25 | 2023-08-18 | 北京嘀嘀无限科技发展有限公司 | 图像处理方法、装置、电子设备及可读存储介质 |
CN109308681B (zh) * | 2018-09-29 | 2023-11-24 | 北京字节跳动网络技术有限公司 | 图像处理方法和装置 |
CN109815788B (zh) * | 2018-12-11 | 2024-05-31 | 平安科技(深圳)有限公司 | 一种图片聚类方法、装置、存储介质及终端设备 |
CN109671016B (zh) * | 2018-12-25 | 2019-12-17 | 网易(杭州)网络有限公司 | 人脸模型的生成方法、装置、存储介质及终端 |
US10937428B2 (en) * | 2019-03-11 | 2021-03-02 | Adobe Inc. | Pose-invariant visual speech recognition using a single view input |
CN110705614A (zh) * | 2019-09-20 | 2020-01-17 | 北京三快在线科技有限公司 | 模型训练方法、装置、电子设备及存储介质 |
CN110650379B (zh) * | 2019-09-26 | 2022-04-01 | 北京达佳互联信息技术有限公司 | 视频摘要生成方法、装置、电子设备及存储介质 |
CN110895557B (zh) * | 2019-11-27 | 2022-06-21 | 广东智媒云图科技股份有限公司 | 基于神经网络的文本特征判断方法、装置和存储介质 |
US11816149B2 (en) * | 2020-02-11 | 2023-11-14 | Samsung Electronics Co., Ltd. | Electronic device and control method thereof |
CN111428767B (zh) * | 2020-03-17 | 2024-03-08 | 深圳市商汤科技有限公司 | 数据处理方法及装置、处理器、电子设备及存储介质 |
CN111507232B (zh) * | 2020-04-10 | 2023-07-21 | 盛景智能科技(嘉兴)有限公司 | 多模态多策略融合的陌生人识别方法和系统 |
CN111738354A (zh) * | 2020-07-20 | 2020-10-02 | 深圳市天和荣科技有限公司 | 一种自动识别训练方法、系统、存储介质及计算机设备 |
US11531697B2 (en) * | 2020-11-03 | 2022-12-20 | Adobe Inc. | Identifying and providing digital images depicting human poses utilizing visual interactive content search and virtual mannequins |
CN114443893A (zh) * | 2021-12-31 | 2022-05-06 | 深圳云天励飞技术股份有限公司 | 人脸数据的归档方法及相关设备 |
Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101515324A (zh) * | 2009-01-21 | 2009-08-26 | 上海银晨智能识别科技有限公司 | 适用于多种姿态的人脸识别布控系统及方法 |
CN101673346A (zh) * | 2008-09-09 | 2010-03-17 | 日电(中国)有限公司 | 对图像进行处理的方法、设备和系统 |
CN102799901A (zh) * | 2012-07-10 | 2012-11-28 | 辉路科技(北京)有限公司 | 一种多角度人脸检测方法 |
CN102831413A (zh) * | 2012-09-11 | 2012-12-19 | 上海中原电子技术工程有限公司 | 基于多分类器融合的人脸识别方法及系统 |
CN103034714A (zh) * | 2012-12-11 | 2013-04-10 | 北京百度网讯科技有限公司 | 移动终端的照片分类管理方法、装置和移动终端 |
CN103177102A (zh) * | 2013-03-22 | 2013-06-26 | 北京小米科技有限责任公司 | 一种图像处理方法和装置 |
US20140320699A1 (en) * | 2013-04-29 | 2014-10-30 | Kuo-Ching Chiang | Photograph Sorting Method for Portable Communication Device |
CN104252618A (zh) * | 2013-06-28 | 2014-12-31 | 广州华多网络科技有限公司 | 提高照片归还速度的方法及系统 |
CN105096377A (zh) * | 2014-05-14 | 2015-11-25 | 华为技术有限公司 | 一种图像处理方法和装置 |
CN105488527A (zh) * | 2015-11-27 | 2016-04-13 | 小米科技有限责任公司 | 图像分类方法及装置 |
Family Cites Families (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR100596559B1 (ko) * | 2003-11-01 | 2006-07-04 | 학교법인연세대학교 | 3차원 얼굴 인식을 위한 오류 보상 특이치 분해 기반 얼굴포즈 추정 방법 |
US7643686B2 (en) * | 2004-11-17 | 2010-01-05 | Eastman Kodak Company | Multi-tiered image clustering by event |
KR100714724B1 (ko) * | 2005-08-17 | 2007-05-07 | 삼성전자주식회사 | 얼굴 포즈 추정 장치와 추정 방법 그리고 상기 방법에 의한얼굴 인식 시스템 |
US9690979B2 (en) * | 2006-03-12 | 2017-06-27 | Google Inc. | Techniques for enabling or establishing the use of face recognition algorithms |
CN101414348A (zh) * | 2007-10-19 | 2009-04-22 | 三星电子株式会社 | 多角度人脸识别方法和系统 |
CN101515234A (zh) | 2009-03-26 | 2009-08-26 | 华亚微电子(上海)有限公司 | 关联数据处理方法和关联数据装置 |
JP5178611B2 (ja) * | 2009-04-15 | 2013-04-10 | 株式会社東芝 | 画像処理装置、画像処理方法及びプログラム |
JP2011019192A (ja) * | 2009-07-10 | 2011-01-27 | Toshiba Corp | 画像表示装置 |
CN101763503B (zh) * | 2009-12-30 | 2012-08-22 | 中国科学院计算技术研究所 | 一种姿态鲁棒的人脸识别方法 |
CN102147851B (zh) * | 2010-02-08 | 2014-06-04 | 株式会社理光 | 多角度特定物体判断设备及多角度特定物体判断方法 |
KR101280439B1 (ko) * | 2011-09-30 | 2013-06-28 | 아이포콤 주식회사 | 현금인출기 카메라에서 취득된 얼굴 영상에 대한 얼굴 인식 가능성 판단 방법 |
CN102722698B (zh) * | 2012-05-17 | 2014-03-12 | 上海中原电子技术工程有限公司 | 多姿态人脸检测与跟踪方法及系统 |
KR20150109987A (ko) * | 2014-03-21 | 2015-10-02 | 삼성전자주식회사 | 영상 처리 장치, 그 제어 방법 및 컴퓨터 판독 가능 기록 매체 |
CN104317932B (zh) * | 2014-10-31 | 2018-04-27 | 小米科技有限责任公司 | 照片分享方法及装置 |
-
2015
- 2015-11-27 CN CN201510846109.1A patent/CN105488527B/zh active Active
-
2016
- 2016-06-29 WO PCT/CN2016/087562 patent/WO2017088470A1/zh active Application Filing
- 2016-11-18 EP EP16199522.0A patent/EP3176727A1/en not_active Ceased
- 2016-11-22 US US15/359,192 patent/US10282597B2/en active Active
Patent Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101673346A (zh) * | 2008-09-09 | 2010-03-17 | 日电(中国)有限公司 | 对图像进行处理的方法、设备和系统 |
CN101515324A (zh) * | 2009-01-21 | 2009-08-26 | 上海银晨智能识别科技有限公司 | 适用于多种姿态的人脸识别布控系统及方法 |
CN102799901A (zh) * | 2012-07-10 | 2012-11-28 | 辉路科技(北京)有限公司 | 一种多角度人脸检测方法 |
CN102831413A (zh) * | 2012-09-11 | 2012-12-19 | 上海中原电子技术工程有限公司 | 基于多分类器融合的人脸识别方法及系统 |
CN103034714A (zh) * | 2012-12-11 | 2013-04-10 | 北京百度网讯科技有限公司 | 移动终端的照片分类管理方法、装置和移动终端 |
CN103177102A (zh) * | 2013-03-22 | 2013-06-26 | 北京小米科技有限责任公司 | 一种图像处理方法和装置 |
US20140320699A1 (en) * | 2013-04-29 | 2014-10-30 | Kuo-Ching Chiang | Photograph Sorting Method for Portable Communication Device |
CN104252618A (zh) * | 2013-06-28 | 2014-12-31 | 广州华多网络科技有限公司 | 提高照片归还速度的方法及系统 |
CN105096377A (zh) * | 2014-05-14 | 2015-11-25 | 华为技术有限公司 | 一种图像处理方法和装置 |
CN105488527A (zh) * | 2015-11-27 | 2016-04-13 | 小米科技有限责任公司 | 图像分类方法及装置 |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110929679A (zh) * | 2019-12-05 | 2020-03-27 | 杭州电子科技大学 | 一种基于gan的无监督自适应行人重识别方法 |
CN110929679B (zh) * | 2019-12-05 | 2023-06-16 | 杭州电子科技大学 | 一种基于gan的无监督自适应行人重识别方法 |
CN111081317A (zh) * | 2019-12-10 | 2020-04-28 | 山东大学 | 一种基于基因谱的乳腺癌淋巴结转移预测方法及预测系统 |
CN111081317B (zh) * | 2019-12-10 | 2023-06-02 | 山东大学 | 一种基于基因谱的乳腺癌淋巴结转移预测方法及预测系统 |
CN112818867A (zh) * | 2021-02-02 | 2021-05-18 | 浙江大华技术股份有限公司 | 人像聚类方法、设备及存储介质 |
CN112818867B (zh) * | 2021-02-02 | 2024-05-31 | 浙江大华技术股份有限公司 | 人像聚类方法、设备及存储介质 |
Also Published As
Publication number | Publication date |
---|---|
CN105488527B (zh) | 2020-01-10 |
US20170154208A1 (en) | 2017-06-01 |
US10282597B2 (en) | 2019-05-07 |
EP3176727A1 (en) | 2017-06-07 |
CN105488527A (zh) | 2016-04-13 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2017088470A1 (zh) | 图像分类方法及装置 | |
US20170154206A1 (en) | Image processing method and apparatus | |
WO2020062969A1 (zh) | 动作识别方法及装置、驾驶员状态分析方法及装置 | |
EP3125158B1 (en) | Method and device for displaying images | |
US10007841B2 (en) | Human face recognition method, apparatus and terminal | |
CN109446994B (zh) | 手势关键点检测方法、装置、电子设备及存储介质 | |
WO2017071085A1 (zh) | 报警方法及装置 | |
WO2017036035A1 (zh) | 屏幕控制方法及装置 | |
WO2017084182A1 (zh) | 图片处理方法及装置 | |
US10115019B2 (en) | Video categorization method and apparatus, and storage medium | |
WO2016011747A1 (zh) | 肤色调整方法和装置 | |
WO2017020476A1 (zh) | 关联用户的确定方法及装置 | |
JP6335289B2 (ja) | 画像フィルタを生成する方法及び装置 | |
WO2017096782A1 (zh) | 防止拍照遮挡的方法及装置 | |
WO2016011745A1 (zh) | 图像分割方法、装置及设备 | |
CN105554389B (zh) | 拍摄方法及装置 | |
CN108668080B (zh) | 提示镜头脏污程度的方法及装置、电子设备 | |
US9924090B2 (en) | Method and device for acquiring iris image | |
CN107944367B (zh) | 人脸关键点检测方法及装置 | |
CN105335684B (zh) | 人脸检测方法及装置 | |
CN105335714B (zh) | 照片处理方法、装置和设备 | |
WO2017140109A1 (zh) | 压力检测方法和装置 | |
WO2016188065A1 (zh) | 云名片推荐方法及装置 | |
CN111259967A (zh) | 图像分类及神经网络训练方法、装置、设备及存储介质 | |
CN106095876B (zh) | 图像处理方法及装置 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 16867697 Country of ref document: EP Kind code of ref document: A1 |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 16867697 Country of ref document: EP Kind code of ref document: A1 |